Google AI Blog: 2020

Reformer: The Efficient Transformer

Thursday, January 16, 2020

Posted by Nikita Kitaev, Student Researcher, UC Berkeley and Łukasz Kaiser, Research Scientist, Google Researchlong short-term memorytranslate sentence-by-sentenceTransformer modelentire Wikipedia articlesgenerate musicimagesattentionthousands of layersReformerlocality-sensitive-hashingreversible residual layersThe Attention Problemhash functionchunks

Locality-sensitive-hashing: Reformer takes in an input sequence of keys, where each key is a vector representing individual words (or pixels, in the case of images) in the first layer and larger contexts in subsequent layers. LSH is applied to the sequence, after which the keys are sorted by their hash and chunked. Attention is applied only within a single chunk and its immediate neighbors.

The Memory Problemgradient descentactivationsreversible layers

Reversible layers: (A) In a standard residual network, the activations from each layer are used to update the inputs into the next layer. (B) In a reversible network, two sets of activations are maintained, only one of which is updated after each layer. (C) This approach enables running the network in reverse in order to recover all intermediate values.

Applications of Reformerthis colab

Top: Image fragments used as input to Reformer. Bottom: “Completed” full-frame images. Original images are from the Imagenet64 dataset.

Crime and Punishmentthis colabConclusionresearch in the openeven longer sequencesimprove handling of positional encodingsReformer paperour codethis colabchat with usAcknowledgementsThis research was conducted by Nikita Kitaev, Łukasz Kaiser and Anselm Levskaya. Additional thanks go to Afroz Mohiuddin, Jonni Kanerva and Piotr Kozakowski for their work on Trax and to the whole JAX team for their support.

Can You Trust Your Model’s Uncertainty?

Wednesday, January 15, 2020

Posted by Jasper Snoek, Research Scientist and Zachary Nado, Research Engineer, Google Researchdeep learningcovariate shiftit was recently observedCan you trust your model’s uncertainty? Evaluating Predictive Uncertainty Under Dataset Shift” NeurIPS 2019What is Out-of-Distribution Data?uncertainty out-of-distribution (OOD)related postQuantifying the Quality of Uncertaintyproper scoring rulesBrier ScoreLog Likelihoodexpected calibration errorExperimentsImageNetCorrupted Imagenet

We explore how model uncertainty behaves under changes to the data distribution, such as increasing intensities of the image perturbations used in Corrupted Imagenet. Shown here are examples of each type of image corruption, at intensity level 3 (of 5).

shifteddropouttemperature scaling

Accuracy (top) and expected calibration error (bottom; lower is better) for increasing intensities of dataset shift on ImageNet-C. We observe that the decrease in accuracy is not reflected by an increase in uncertainty of the model, indicated by both accuracy and ECE getting worse.

temperature scalingPlatt scalingDeep ensemblesSummary and Recommended Best Practices

Uncertainty under dataset shift is a real concern that needs to be considered when training models.

Improving calibration and accuracy on an in-distribution test set often does not translate to improved calibration on shifted data.

Out of all the methods we considered, deep ensembles are the most robust to dataset shift, and a relatively small ensemble size (e.g., 5) is sufficient. The effectiveness of ensembles presents interesting avenues for improving other approaches.

code and model predictions

Using Machine Learning to “Nowcast” Precipitation in High Resolution

Monday, January 13, 2020

Posted by Jason Hickey, Senior Software Engineer, Google ResearchMachine Learning for Precipitation Nowcasting from Radar Imagesprecipitation nowcasting, Moving Beyond Traditional Weather ForecastingDoppler radar

Top: Image showing the location of clouds as measured by geosynchronous satellites. Bottom: Radar image showing the location of rain as measured by Doppler radar stations. (Credit: NOAA, NWS, NSSL)

National Oceanic and Atmospheric Administrationright nownowcastingRadar-to-Radar ForecastingNNadvectionconvection

Top (left to right): The first three panels show radar images from 60 minutes, 30 minutes, and 0 minutes before now, the point at which a prediction is desired. The right-most panel shows the radar image 60 minutes after now, i.e., the ground truth for a nowcasting prediction. Bottom Left: For comparison, a vector field induced from applying an optical flow (OF) algorithm for modeling advection to the data from the first three panels above. Optical flow is a computer vision method that was developed in the 1940s, and is frequently used to predict short term weather evolution. Bottom Right: An example prediction made by OF. Notice that it tracks the motion of the precipitation in the bottom left corner well, but fails to account for the decaying strength of the storm.

physics-freea priorilayersconvolutional filtersU-Netencodingdecodingphase

(A) The overall structure of our U-NET. Blue boxes correspond to basic CNN layers. Pink boxes correspond to down-sample layers. Green boxes correspond to up-sample layers. Solid lines indicate input connections between layers. Dashed lines indicate long skip connections transversing the encoding and decoding phases of the U-NET. Dotted lines indicate short skip connections for individual layers. (B) The operations within our basic layer. (C) The operations within our down-sample layers. (D) The operations within our up-sample layers.

NResultsHigh Resolution Rapid Refresh1-hour total accumulated surface precipitationpersistence modelnow

A visualization of predictions made over the course of roughly one day. Left: The 1-hour HRRR prediction made at the top of each hour, the limit to how often HRRR provides predictions. Center: The ground truth, i.e., what we are trying to predict. Right: The predictions made by our model. Our predictions are every 2 minutes (displayed here every 15 minutes) at roughly 10 times the spatial resolution made by HRRR. Notice that we capture the general motion and general shape of the storm.

precision and recall

Precision and recall (PR) curves comparing our results (solid blue line) with: optical flow (OF), the persistence model, and the HRRR 1-hour prediction. As we do not have direct access to their classifiers, we cannot provide a full PR curve for their results. Left: Predictions for light rain. Right: Predictions for moderate rain.

AcknowledgementsThanks to Carla Bromberg, Shreya Agrawal, Cenk Gazen, John Burge, Luke Barrington, Aaron Bell, Anand Babu, Stephan Hoyer, Lak Lakshmanan, Brian Williams, Casper Sønderby, Nal Kalchbrenner, Avital Oliver, Tim Salimans, Mostafa Dehghani, Jonathan Heek, Lasse Espeholt, Sella Nevo, Avinatan Hassidim.

Google Research: Looking Back at 2019, and Forward to 2020 and Beyond

Thursday, January 9, 2020

Posted by Jeff Dean, Senior Fellow and SVP of Google Research and Health, on behalf of the entire Google Research communityGoogle Research201820172016research publications in 2019Ethical Use of AIAI Principlesone-year update

Published a research paper about a new transparency tool, which enabled the launch of Model Cards for several of our Cloud AI products. You can see an example model card for the Cloud AI Vision API Object Detection feature.

Showed how Activation Atlases can help explore neural network behavior and can aid with interpretability of machine learning models.

Introduced TensorFlow Privacy, an open-source library to enable training machine learning models with differential privacy guarantees.

Released a beta version of Fairness Indicators, to help ML practitioners identify unjust or unintended impacts of machine learning models.

Clicking on a slice in Fairness Indicators will load all the data points in that slice inside the What-If Tool widget. In this case, all data points with the “female” label are shown.

Published a KDD'19 paper on how pairwise comparisons and regularization is incorporated into a large-scale production recommender system to improve ML Fairness.

Published an AIES'19 paper about a case study on the application of fairness in machine learning research to a production classification system, and described our fairness metric, conditional equality, that takes into account distributional differences in implementing equality of opportunity.

Published an AIES'19 paper about counterfactual fairness in text classification problems that asks the question: "How would the prediction change if the sensitive attribute referenced in the example were different?" and used this approach to improve our production systems that assess the toxicity of online content.

Released a new dataset to help with research to identify deepfakes.

A sample of videos from Google’s contribution to the FaceForensics benchmark. To generate these, pairs of actors were selected randomly, and deep neural networks swapped the face of one actor onto the head of another.

AI for Social GoodFloods are the most common and the most deadly natural disaster on the planetmachine learning, computation and better sources of data to make significantly more accurate flood forecastsworkshop that brought together researchers with expertise in flood forecasting, hydrology and machine learning

use machine learning to help analyze wildlife camera dataNOAAidentify whale species and locationscreated and released a set of tools for enabling new kinds of machine-learning-oriented biodiversity research6th Fine-Grained Visual Categorization WorkshopMakerere University AI & Data Science research groupcassava plant diseasesGoogle Earth Timelapseaggregate data on human mobility617 million children do not have basic literacyBolo app uses speech-recognition technologyhelped 800,000 children read storiesthree-month pilot

Socratic apphelp high schoolersSocratic methodannouncedAI Impact Challenge

The Fondation Médecins Sans Frontières (MSF) is creating a free smartphone application that uses image recognition tools to help clinical staff in low-resource settings (currently being piloted in Jordan) to analyze anti-microbial images and advise on the appropriate antibiotics to use for a particular patient’s infection.

Over a billion people live in smallholder farm households. A single pest attack can devastate their crop yields and livelihoods. Wadhwani AI uses image classification models that can identify pests and provide timely advice on what pesticides to spray and when—ultimately improving crop yield.

And deep in tropical rainforests, where illegal deforestation is a major driver of climate change, Rainforest Connection uses deep learning for bioacoustic monitoring and old cell phones to track rainforest health and detect threats.

Our 20 AI Impact Challenge winners. You can learn more about the work of all the grantees here.

Applications of AI to Other Fieldshave published a number of papers in

In An Interactive, Automated 3D Reconstruction of a Fly Brain, we reported on a collaborative effort that achieved a milestone of mapping the structure of an entire fly brain, using machine learning models that were able to painstakingly trace each individual neuron.

In Learning Better Simulation Methods for Partial Differential Equations (PDEs), we showed how machine learning can be used to accelerate PDE computations, which are at the heart of many fundamental computational problems in climate science, fluid dynamics, electromagnetism, heat conduction and general relativity.

Simulations of Burgers’ equation, a model for shock waves in fluids, solved with either a standard finite volume method (left) or our neural network based method (right). The orange squares represent simulations with each method on low resolution grids. These points are fed back into the model at each time step, which then predicts how they should change. Blue lines show the exact simulations used for training. The neural network solution is much better, even on a 4x coarser grid, as indicated by the orange squares smoothly tracing the blue line.

We gave machine learning models better scents of the world with Learning to Smell: Using Deep Learning to Predict the Olfactory Properties of Molecules. We showed how to leverage graph neural networks (GNNs) to directly predict the odor descriptors for individual molecules, without using any handcrafted rules.

2D snapshot of our embedding space with some example odors highlighted. Left: Each odor is clustered in its own space. Right: The hierarchical nature of the odor descriptor. Shaded and contoured areas are computed with a kernel-density estimate of the embeddings.

In work that combines chemistry and reinforcement learning techniques, we presented a framework for molecule optimization.

Machine learning can also help us in our artistic and creative endeavors. Artists have found ways to collaborate with AI and AR and create interesting new forms, from dancing with a machine to reimagine choreography, to creating new melodies with machine learning tools. ML can be used by novices, too. To honor the birthday of J.S. Bach, we featured a ML-powered Doodle: just create your melody, and the ML tool can create accompanying harmonizations in Bach’s style.

Assistive Technology

Lookout helps people who are blind or have low vision identify information about their surroundings. It draws upon similar underlying technology as Google Lens, which lets you search and take action on the objects around you, simply by pointing your phone.

Live Transcribe has the potential to give people who are deaf or hard of hearing greater independence in their everyday interactions. You can get real-time transcriptions of conversations that the user is engaged in, even if the speech is in another language.

Project Euphonia performs personalized speech-to-text transcription. For people with ALS and other conditions that produce slurred or non-standard speech, this research improves automatic speech recognition (ASR) over other state-of-the-art ASR models.

Like Project Euphonia, Parrotron uses end-to-end neural networks to help improve communication, but the research focuses on automatic speech-to-speech conversion rather than transcription, presenting a speech interface that may be easier for some to access.

Millions of images online don’t have any text description. Get Image Descriptions from Google helps blind or low vision users understand unlabelled images. When a screen reader encounters an image or graphic without a description, Chrome can now create one automatically.

We developed tools that can read visual text in audio form in Lens for Google Go, greatly helping users who are not fully literate navigate the word-rich world around them.

Making Your Phone More Intelligentspeech recognition modelsvision modelshandwriting recognition models

The launch of on-device captioning with Live Caption, giving always-available transcription of any video playing on your device.

The creation of a powerful new transcribing Recorder app, which can help index audio information and make it easily retrievable.

Improvements to Google Translate’s camera translation, so that you can point at text in an unfamiliar language and get it instantly translated in context.

Release of the Augmented Faces API in ARCore, enabling new real-time AR self-expression tools.

A demonstration of on-device, real-time hand tracking, enabling new ways for users to interact with and control their devices with their hands.

Improved, RNN-based on-device handwriting recognition for on-screen mobile keyboards.

The release of a new global localization approach using your smart phone’s camera to help more accurately orient you and help you find your way in the world.

Federated learningonline comicpowerful machine learning approach invented by Google researchers in 2015large-scale learning systemssurvey article on Federated Learningtake great selfiesprofessional-looking shallow depth of field images and portraitsthe Night Sight feature on Pixel Phones to take some stunning astrophotography picturesmulti-frame super resolutionmobile photography in very low-light conditions

HealthGoogle Healthpublishing research papersbuilding tools

We showed that a deep learning model for mammography can assist physicians in spotting breast cancer, a condition that affects 1 in 8 women in the US during their lifetimes, with greater accuracy than experts, reducing both false positives and false negatives. The model trained on de-identified data from a UK hospital had similar gains in accuracy when used to evaluate patients in a completely different healthcare system in the U.S.

Example of a difficult-to-detect cancer case correctly identified by machine learning.

We showed that a deep learning model for differential diagnoses of skin diseases can give results that are significantly more accurate than primary care physicians and on par with or perhaps slightly better than dermatologists.

Working alongside experts from the US Department of Veterans Affairs (VA), DeepMind Health colleagues who are now part of Google Health showed that a machine learning model can predict the onset of acute kidney injury (AKI), one of the leading causes of avoidable patient harm, up to two days before it happens. In the future, this could give doctors a 48-hour head start in treating this serious condition.

We expanded the application of deep learning to electronic health records with several partner organizations. You can read more about this work in our 2018 blog post.

We showed a promising step forward for predicting lung cancer, where a deep learning model for examining the results of a single CT scan study performed on par or better than trained radiologists at early detection of lung cancer. Early detection of lung cancer dramatically improves survival rates.

We continued to expand and evaluate our deployment of machine learning tools for detection and prevention of eye disease, in collaboration with Verily and our healthcare partners in India and Thailand.

We published a research paper on an augmented reality microscope for cancer diagnosis, whereby a pathologist can get real-time feedback about what parts of a slide are most interesting while examining tissue through a microscope. You can also read more about it in our 2018 blog post here.

We built a human-centric, similar-image search tool for pathologists to help them make more effective diagnoses, by allowing examination of similar cases.

Quantum Computingcomputational task

Left: Artist's rendition of the Sycamore processor mounted in the cryostat. (Full Res Version; Forest Stearns, Google AI Quantum Artist in Residence) Right: Photograph of the Sycamore processor. (Full Res Version; Erik Lucero, Research Scientist and Lead Production Quantum Hardware)

early exampleeasier to expresseasier to controlclassical machine learning techniques like deep reinforcement learningwhat our quantum computing milestone meansGeneral Algorithms and Theoryalgorithms and theorygraph miningmarket algorithmssummarizing some of our work in graph learning algorithmsVLDB’19Cache-aware load balancing of data center applicationsbalanced partitioning of graphs

Heatmap of flash IO requests (resulting from cache misses) across web search serving leaves. The three humps represent random leaf selection, load balancing, and cache-aware load balancing (left to right). Lines indicate the 50th, 90th, 95th and 99.9th percentiles. From VLDB’19 paper, "Cache-aware load balancing of data center applications."

ICLR’2019A new dog learns old tricks: RL finds classic optimization algorithmsFOCS’19 papertheorypracticedensity clusteringvocabulary compressionSODA’19 paperFOCS 2019 paperbipartite matchingITCS’19cachingSODA’20market algorithmsinnovations in experimental designNeurIPS’19 oral paperWINE2019KDD'19 paperNeurIPS'19 paperexperimental power

The clustering algorithm from the KDD’19 paper “Randomized Experimental Design via Geographic Clustering“ applied to user queries from the United States. The algorithm automatically identifies metropolitan areas, correctly predicting, for example, that the Bay Area includes San Francisco, Berkeley, and Palo Alto, but not Sacramento.

Machine Learning AlgorithmsMeasuring the Limits of Data Parallel Training for Neural Networkspaper

For all workloads we tested, we observed a universal relationship between batch size and training speed with three distinct regimes: perfect scaling with small batch sizes (following the dashed line), eventually seeing diminishing returns as batch sizes grow (diverging from the dashed line), and maximal data parallelism at the largest batch sizes (where the trend plateaus). The transition points between the regimes vary dramatically between different workloads.

GPipe

In Evaluating the Unsupervised Learning of Disentangled Representations, we examined what properties affect the representations that are learned from unsupervised data, in order to better understand what makes for good representations and effective learning.

In Predicting the Generalization Gap in Deep Neural Networks, we showed that it is possible to predict the generalization gap (the gap between a model’s performance on data from the training distribution versus data drawn from a different distribution) using statistics of the margin distribution, helping us better understand which models generalize most effectively. We also did some research on Improving Out-of-Distribution Detection in Machine Learning Models, to better understand when a model is starting to encounter kinds of data it has never seen before. We also looked at Off-Policy Classification in the context of reinforcement learning, to better understand which models are likely to generalize the best.

In Learning to Generalize from Sparse and Underspecified Rewards, we also examined ways of specifying reward functions for reinforcement learning that enable learning systems to more directly learn from true objectives and be less distracted with longer, less-desirable sequences of actions that happen to achieve desired goals by accident.

In this instruction-following task, the action trajectories a₁, a₂ and a₃ reach the goal, but the sequences a₂ and a₃ do not follow the instructions. This illustrates the issue of underspecified rewards.

AutoML

In EfficientNet: Improving Accuracy and Efficiency through AutoML and Model Scaling, we showed how to use neural architecture search techniques to achieve substantially better results on computer vision problems, including a new state-of-the-art result of 84.4% top-1 accuracy on ImageNet while having 8X fewer parameters than the previous best model.

Model Size vs. Accuracy Comparison. EfficientNet-B0 is the baseline network developed by AutoML MNAS, while Efficient-B1 to B7 are obtained by scaling up the baseline network. In particular, our EfficientNet-B7 achieves new state-of-the-art 84.4% top-1 / 97.1% top-5 accuracy, while being 8.4x smaller than the best existing CNN.

In EfficientNet-EdgeTPU: Creating Accelerator-Optimized Neural Networks with AutoML, we showed how a neural architecture search approach can find efficient models that are tailored to particular hardware accelerators, resulting in high accuracy, low-computational models for running on mobile devices.

In Video Architecture Search, we describe how we extended our AutoML work to the domain of video models, finding architectures that achieve state-of-the-art results, and also lightweight architectures that match the performance of hand-crafted models while using 50x less computation.

TinyVideoNet (TVN) architectures evolved to maximize the recognition performance while keeping its computation time within the desired limit. For instance, TVN-1 (top) runs at 37 ms on a CPU and 10ms on a GPU. TVN-2 (bottom) runs at 65ms on a CPU and 13ms on a GPU.

We developed AutoML techniques for tabular data, unlocking an important domain where many companies and organizations have interesting data in relational databases, and often want to develop machine learning models on this data. We collaborated to release this technology as a new Google Cloud AutoML Tables product, and also discussed how well this system did in a new Kaggle competition in An End-to-End AutoML Solution for Tabular Data at KaggleDays (spoiler: AutoML Tables finished second out of 74 teams of expert data scientists).

In Exploring Weight Agnostic Neural Networks, we showed how it is possible to find interesting neural network architectures without any training steps to update the weights of the evaluated models. This can make architecture search much more computationally efficient.

A weight-agnostic neural network performing a Cartpole Swing-up task at various different weight parameters, and also using fine-tuned weight parameters.

Applying AutoML to Transformer Architectures explored finding architectures for natural language processing tasks that significantly outperform vanilla Transformer models at substantially reduced computational costs.

Comparison between the Evolved Transformer and the original Transformer on WMT’14 En-De at varying sizes. The biggest gains in performance occur at smaller sizes, while ET also shows strength at larger sizes, outperforming the largest Transformer with 37.6% less parameters (models to compare are circled in green). See Table 3 in our paper for the exact numbers.

In SpecAugment: A New Data Augmentation Method for Automatic Speech Recognition, we showed that the approach of automatically learning data augmentation methods can be extended to speech recognition models, with the learned augmentation approaches achieving significantly higher accuracy with less data than existing human ML-expert driven data augmentation approaches.

We launched our first speech application for keyword spotting and spoken language identification using AutoML. In our experiments we found better models (both more efficient and better performance) than the human designed models that have been in this setting for some time.

Natural Language Understanding

In Exploring Massively Multilingual, Massive Neural Machine Translation, we showed significant gains in translation quality by training a single model to translate between 100 languages, rather than having 100 separate models.

Left: Language pairs with larger amounts of training data generally have higher translation quality. Right: Multilingual training, where we train a single model for all language pairs rather than separate models for each language pair, results in substantial improvements in BLEU score (a measure of translation quality) for language pairs without much data.

In Large-Scale Multilingual Speech Recognition with a Streaming End-to-End Model, we showed how combining speech recognition and language models together and training the system on many languages, can significantly improve speech recognition accuracy.

Left: A traditional monolingual speech recognizer comprised of Acoustic, Pronunciation and Language Models for each language. Middle: A traditional multilingual speech recognizer where the Acoustic and Pronunciation model is multilingual, while the Language model is language-specific. Right: An E2E multilingual speech recognizer where the Acoustic, Pronunciation and Language Model is combined into a single multilingual model.

In Translatotron: An End-to-End Speech-to-Speech Translation Model, we showed that it is possible to train a joint model to accomplish the (normally separate) tasks of speech recognition, translation and text-to-speech generation with nice benefits, like preserving the sound of the speaker’s voice in the generated translated audio, as well as a simpler overall learning system.

In Multilingual Universal Sentence Encoder for Semantic Retrieval, we showed how to combine many different objectives to yield models that are significantly better at semantic retrieval (versus simpler word matching techniques). For example, in Google Talk to Books, the query “What fragrance brings back memories?” yields the result, “And for me, the smell of jasmine along with the pan bagnat, it brings back my entire carefree childhood.”

In Robust Neural Machine Translation, we showed how to use an adversarial training procedure to significantly improve the quality and robustness of language translations.

Left: The Transformer model is applied to an input sentence (lower left) and, in conjunction with the target output sentence (above right) and target input sentence (middle right; beginning with the placeholder “<sos>”), the translation loss is calculated. The AdvGen function then takes the source sentence, word selection distribution, word candidates and the translation loss as inputs to construct an adversarial source example. Right: In the defense stage, the adversarial source example serves as input to the Transformer model and the translation loss is calculated. AdvGen then uses the same method as above to generate an adversarial target example from the target input.

seq2seqTransformerBERTTransformer-XLALBERTGoogle TranslateSmart ComposeGoogle Searchlaunch of BERT in our core search and ranking algorithmsMachine Perception

Finer-grained visual understanding in Lens, enabling even more powerful visual search.

Helpful smart camera features such as Quick Gestures, Face Match and smart video call framing on the Nest Hub Max.

Technology for live and spatially-aware perception for helpfully augmenting the world around us through Lens.

Better models for depth prediction from videos.

Better representations for fine-grained temporal understanding of videos using temporal cycle-consistency learning.

Right: Input videos of people performing a squat exercise. The video on the top left is the reference. The other videos show nearest neighbor frames (in the TCC embedding space) from other videos of people doing squats. Left: The corresponding frame embeddings move as the action is performed.

Learning representations across text, speech and video that are temporally consistent from unlabeled videos.

Qualitative results from VideoBERT, pretrained on cooking videos. Top: Given some recipe text, we generate a sequence of visual tokens. Bottom: Given a visual token, we show the top three future tokens forecast by VideoBERT at different time scales. In this case, the model predicts that a bowl of flour and cocoa powder may be baked in an oven and may become a brownie or cupcake. We visualize the visual tokens using the images from the training set closest to the tokens in feature space.

Being able to predict future visual inputs from observations of the past.

Models that can better understand action sequences in videos, enabling you to better recall special video moments like “blowing out candles” or “sliding down a slide” in Google Photos.

Architecture for temporal action localization.

Roboticssignificant research area

In Long-Range Robotic Navigation via Automated Reinforcement Learning, we showed how to combine reinforcement learning with long-range planning to enable robots to more effectively navigate complex environments (like our Google office buildings).

In PlaNet: A Deep Planning Network for Reinforcement Learning, we showed how to effectively learn a world model purely from the pixels of images, and how to leverage this model of how the world behaves in order to accomplish tasks with many fewer learning episodes.

In Unifying Physics and Deep Learning with TossingBot, we showed how robots can learn “intuitive” physics from experimentation in an environment, rather than being pre-programmed with physics models about the environment in which they are operating.

In Soft Actor-Critic: Deep Reinforcement Learning for Robotics, we showed that training a reinforcement learning algorithm to both maximize the expected reward (which is the standard RL objective) and to maximize the policy's entropy (so that learning favors policies that are more random), can help robots learn faster and be more robust to changes in their environment.

In Learning to Assemble and to Generalize from Self-Supervised Disassembly, we showed how robots can learn to assemble by first learning to disassemble things in a self-supervised way. Kids learn from taking things apart, and it appears that robots can as well!

We introduced ROBEL: Robotics Benchmarks for Learning with Low-Cost Robots, an open-source platform of cost-effective robots and curated benchmarks designed to facilitate research and development on physical robotics hardware in the real world.

Helping Advance the Broader Developer and Researcher Communitylaunched TensorFlow 2.0mobile GPU inferenceTensorFlow Litelaunched Teachable Machine 2.0MLIRJAXNeurIPS 2019neural tangent kernelsBayesian inferencemolecular dynamicsa preview of JAX on Cloud TPUsMediaPipeXNNPACKCloud TPUsTensorFlow Research CloudIntro To TensorFlow at Courserataking TensorFlow on the roadTensorFlow Worlddiscovered two new planetsreminiscent of African maskscreate the Farmers Companiondetermine safe road conditionsidentify pot holes and dangerous road crackslearns how to add color to black-and-white photosOpen DatasetsGoogle Dataset Searchshare open data responsibly

Open Images V5: An update to the popular Open Images dataset that includes segmentation masks for 2.8 million objects in 350 categories (so that it now has ~9M images annotated with image-level labels, object bounding boxes, object segmentation masks, and visual relationships).

Natural questions: the first dataset to use naturally occurring queries and find answers by reading an entire page, rather than extracting answers from a short paragraph.

Data for deepfake detection: we contributed a large dataset of visual deepfakes to the FaceForensics benchmark (mentioned above).

Google Research Football: a novel reinforcement learning environment where agents aim to master the world’s most popular sport—football (or, if you’re American, soccer). It’s important for reinforcement learning agents to have GOOOAAALLLSS!

Google-Landmarks-v2: over 5 million images (2x that of the first release) of more than 200 thousand different landmarks.

YouTube-8M Segments: A large-scale classification and temporal localization dataset that includes human-verified labels at the 5-second segment level of YouTube-8M videos.

Atomic Visual Actions (AVA) Spoken Activity: A multimodal audio+visual video dataset for perception of conversations. In addition, academic challenges were run for AVA action recognition and AVA: Spoken Activity

PAWS and PAWS-X: To help with paraphrase identification, both datasets contain well-formed sentence pairs with high lexical overlap, in which around half of pairs are paraphrase and half are not.

Natural language dialog datasets: CCPE and Taskmaster-1 both use a Wizard-of-Oz platform that pairs two people who engage in spoken conversations, to mimic a human-level conversation with a digital assistant.

The Visual Task Adaptation Benchmark: VTAB follows similar guidelines to ImageNet and GLUE but is based on one principle—a better representation is one that yields better performance on unseen tasks, with limited in-domain data.

Schema-Guided Dialogue Dataset: the largest publicly available corpus of task-oriented dialogues, with over 18,000 dialogues spanning 17 domains.

Research Community Interaction

CVPR: ~250 Googlers presented 40+ papers, talks, posters, workshops and more.

ICML: ~200 Googlers presented 100+ papers, talks, posters, workshops and more.

ICLR: ~200 Googlers presented 60+ papers, talks, posters, workshops and more.

ACL: ~100 Googlers presented 40+ papers, workshops and tutorials.

Interspeech: Over 100 Googlers presented 30+ papers.

ICCV: ~200 Googlers presented 40+ papers, and several Googlers also won three prestigious ICCV awards.

NeurIPS: ~500 Googlers co-authored more than 120 accepted papers and engaged in various workshops and more.

Google Faculty Research Awards 2018Google AI Residency Programmentored AI-focused startupsNew Places, New FacesResearch office in Bangalorewe’re hiringLooking Forward to 2020 and Beyonda nice overview of important advances of the last decadeTPUv1TPUv2 and TPUv3Edge TPUsdeep learning revolution will continue to reshape how we think about computing and computers

How can we build machine learning systems that can handle millions of tasks, and that can learn to successfully accomplish new tasks automatically? Currently, we’re mostly training separate machine models for each new task, starting from scratch, or at best, from a model trained on one or a few highly related tasks. As such, the models we train are really good at one or a few things, but not good at anything else. However, what we truly want are models that are good at leveraging their expertise at doing many things, so that they are able to learn to do a new thing with relatively little training data and computation. This is a true grand challenge which will require expertise and advances in many areas spanning solid-state circuit design, computer architecture, ML-focused compilers, distributed systems, machine learning algorithms and domain experts across many other fields in order to build systems that can generalize to solve new tasks independently across a full range of application areas.

How can we advance the state-of-the-art in important areas of artificial intelligence research like avoiding bias, increasing interpretability & understandability, improving privacy and ensuring safety? Advances in these areas are going to be critical as we use machine learning in more and more ways in society.

How can we apply computation and machine learning to make advances in important new areas of science? There are important advances to be had by collaborating with experts in other fields in areas like climate science, healthcare, bioinformatics and many other areas.

How can we ensure that the ideas and directions pursued by the machine learning and computer science research communities are put forth and explored by a diverse group of researchers? The work that the computer science and machine learning research communities are pursuing has broad implications for billions of people, and we want the set of researchers doing this work to represent the experiences, perspectives, concerns and creative enthusiasm of all the people of the world. How can we best support new researchers from diverse backgrounds entering the field?